Understanding the Relationship Between Social Anxiety and Alcohol Use in College Students: a Meta-Analysis

Addictive Behaviors 38 (2013) 2690–2706

Contents lists available at ScienceDirect

Addictive Behaviors

Review Understanding the relationship between social anxiety and alcohol use in college students: A meta-analysis

Amie R. Schry ⁎, Susan W. White

Virginia Tech, Department of Psychology, 109 Williams Hall (0436), Blacksburg, VA 24061, United States

HIGHLIGHTS

• Meta-analysis examining social anxiety and alcohol variables in college students. • Social anxiety (SA) was negatively related to alcohol use. • SA was positively related to alcohol-related problems. • Drinking motives and alcohol outcome expectancies were positively related to SA. • Clinicians should be aware of these relationships for treatment and assessment.

article info abstract

Keywords: Many college students use alcohol, and most of these students experience problems related to their Social anxiety use. Emerging research indicates that socially anxious students face heightened risk of experiencing Alcohol alcohol-related problems, althoughtheextantresearchonalcoholuseandsocialanxietyinthispopula- Undergraduates tion has yielded inconsistent findings. This meta-analysis was conducted to examine the relationship Meta-analysis between social anxiety and alcohol variables in college students. A literature search was used to identify studies on college students that included measures of social anxiety and at least one of the alcohol variables of interest. All analyses were conducted using random effects models. We found that social anxiety was negatively correlated with alcohol use variables (e.g., typical quantity and typical frequency), but significantly positively correlated with alcohol-related problems, coping, conformity, and social motives for alcohol use, and positive and negative alcohol outcome expectancies. Several moderators of effect sizes were found to be significant, including methodological factors such as sample ascertainment approach. Given that social anxiety was negatively related to alcohol use but positively related to alcohol-related problems, research is needed to address why individuals high in social anxiety experience more problems as a result of their alcohol use. Avoidance of social situations among socially anxious students should also be taken into account when measuring alcohol use. The primary limitation of this study is the small number of studies available for inclusion in some of the analyses. © 2013 Elsevier Ltd. All rights reserved.

Contents

1. Introduction ...... 2691 1.1. Social anxiety and alcohol ...... 2691 1.2. Potential mediators: Drinking motives and alcohol outcome expectancies ...... 2692 1.3. Potential moderators ...... 2692 1.4. Hypotheses ...... 2693 2. Materials and methods ...... 2693 2.1. Operational deﬁnitions ...... 2693 2.2. Literature search ...... 2693

⁎ Corresponding author at: Department of Psychology, Virginia Tech, 109 Williams Hall (0436), Blacksburg, VA 24061, United States. Tel.: +1 540 231 6744; fax: +1 540 231 3652. E-mail addresses: [email protected] (A.R. Schry), [email protected] (S.W. White).

2.3. Study inclusion criteria and resulting pool of studies ...... 2693 2.4. Study coding ...... 2694 2.4.1. Categorical moderators ...... 2694 2.4.2. Continuous moderators ...... 2696 2.5. Effect sizes ...... 2697 2.6. Statistical analyses ...... 2697 3. Results ...... 2697 3.1. Mean effect sizes and homogeneity analyses ...... 2697 3.2. Analyses of moderator variables...... 2697 4. Discussion ...... 2698 4.1. Limitations ...... 2700 4.2. Recommendations for future research ...... 2700 4.3. Clinical implications ...... 2700 4.4. Conclusions ...... 2700 Role of funding sources ...... 2700 Contributors ...... 2700 Conﬂict of interest ...... 2700 Acknowledgments ...... 2700 Appendix A...... 27002701 References ...... 2705

1. Introduction dampening effect of alcohol on social anxiety and because college students are likely to be in situations where drinking, often to Alcohol is, by far, the most widely used drug among college excess, is the social norm, understanding relationships among students, with 60.8% of students reporting alcohol use in the past social anxiety and alcohol-related variables may be particularly month (Substance Abuse & Mental Health Services Administration, relevant. 2012). In 2009, 61.5% of college students reported that they had Social anxiety disorder (SAD) and alcohol use disorders (AUDs) been intoxicated at least once in the past year, with 42.4% reporting are frequently comorbid (see Morris, Stewart, and Ham (2005), that they had been intoxicated in the past 30 days (Johnston, for a review). Approximately 13% of adults with past-year SAD O'Malley, Bachman, & Schulenberg, 2010). Approximately 37 to met criteria for a comorbid AUD, and of adults with lifetime SAD, 44% of college students reported that they binge drank at least 48.2% met criteria for an AUD (Grant et al., 2005). This relationship once in the past two weeks to month (Hingson, Heeren, Winter, & appears to be due to a greater likelihood of having comorbid alcohol Wechsler, 2005; Johnston et al., 2010; Wechsler et al., 2002). dependence (characterized by tolerance, withdrawal, or compulsive Alcohol use can lead to a wide range of problems (e.g., involve- alcohol consumption (APA, 2000); OR = 2.26 to 2.7) rather than ment in risky sexual situations, driving under the influence, hang- alcohol abuse (characterized by a pattern of negative consequences overs, nausea and vomiting, and aggression). Due to the high that result from alcohol use (APA, 2000); OR = 1.2 to 1.23; Buckner, levels of alcohol consumption and the contexts in which college Timpano, Zvolensky, Sachs-Ericsson, & Schmidt, 2008; Grant et al., students typically consume alcohol (e.g., parties where excessive 2005). Both retrospective and longitudinal studies have shown drinking is the norm), along with no parental oversight and moni- that when SAD and AUD co-occur, SAD typically precedes the onset toring, this population may be particularly likely to experience of the AUD (Buckner, Schmidt, et al., 2008; Buckner, Timpano, et al., alcohol-related problems (ARPs). Of college students who drank 2008; Buckner & Turner, 2009; Falk, Yi, & Hilton, 2008). at least once per week during their first year of college, 80% ex- Consistent with studies of adults, Kushner and Sher (1993) found perienced more than one ARP during their first year, and 34% that 43% of college freshmen with SAD met diagnostic criteria for an reported that they had experienced six or more ARPs during that AUD while only 26% of college freshman without SAD met criteria time (Mallett et al., 2011). for an AUD. Overall, however, research on the relationship between social anxiety and alcohol use among college students has revealed 1.1. Social anxiety and alcohol very mixed findings (see Morris et al. (2005), for a review). Some laboratory studies have demonstrated that socially anxious par- It is important for practitioners who work with this population ticipants drink more in anticipation of both interaction (Higgins to be aware of other problems that may be associated with, or lead & Marlatt, 1975) and speech tasks (Kidorf & Lang, 1999), whereas to, problematic alcohol use in order to best treat students, even others (e.g., Holroyd, 1978) have found that socially anxious students when ARPs are not the primary presenting concern. Social anxiety, drink significantly less alcohol than non-socially anxious peers during defined as a fear of being observed or negatively evaluated by others informal laboratory-based “get togethers.” Survey studies of college (American Psychiatric Association [APA], 2000), has been studied students have either failed to find a relationship between social in relation to alcohol use among college students. For individuals anxiety and alcohol consumption, or have found an inverse relation- with high levels of social anxiety, fear of negative evaluation ship between social anxiety and alcohol consumption (e.g., Buckner, may play a role in their alcohol use, as they may “use substances Schmidt, & Eggleston, 2006; Gilles, Turk, & Fresco, 2006; Ham & to avoid potential scrutiny from substance-using peers and/or Hope, 2006; Lewis et al., 2008). One possible reason for the lack because they believe substance use is a common (and thus socially of a positive relationship between social anxiety and alcohol use acceptable) strategy for dampening their anxiety in social situa- is that socially anxious students may avoid social situations and tions” (Buckner, 2011, p. 4). Multiple theoretical models (e.g., only use alcohol to cope with anxiety in social situations when they Appraisal-Disruption Model (Sayette, 1993), Attention Allocation cannot be avoided (Norberg, Norton, & Olivier, 2009; Stewart, Model (Steele, Southwick, & Pagano, 1986), and Avoidance-Coping Morris, Mellings, & Komar, 2006). Model (Bacon & Ham, 2010)) suggest that alcohol use can result Despite the fact that many studies have found a negative relation- in decreased anxiety for at least some individuals. In light of the ship, or no relationship at all, between social anxiety and alcohol use, 2692 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 many studies have found that social anxiety is positively associat- enhancement motives mediated the relationship between social anxi- ed with ARPs (e.g., Buckner, Ecker, & Proctor, 2011; Buckner & ety and ARPs (Buckner et al., 2006). Heimberg, 2010; Buckner et al., 2006; Gilles et al., 2006; Norberg A number of different AOEs have been assessed in the literature et al., 2009). A significant relationship between social anxiety and (e.g., tension reduction, sociability, and impairment). Although ARPs may be particularly important, because AUDs are defined there are a variety of expectancies, they can all be classified as either by problems resulting from the use of alcohol rather than simply positive or negative, depending on whether it would be a desirable the quantity and frequency of use (Buckner et al., 2006). However, or undesirable outcome (Jones et al., 2001). Social anxiety may not all studies have found a significant relationship between social be related to both positive and negative AOEs, as socially anxious in- anxiety and ARPs (e.g., Ham, Zamboanga, Bacon, & Garcia, 2009; dividuals may drink because they expect certain positive outcomes LaBrie, Pedersen, Neighbors, & Hummer, 2008). (e.g., being more sociable) but not drink to excess due to fears about negative outcomes (e.g., doing something embarrassing out of impairment from alcohol). AOEs have been shown to predict al- 1.2. Potential mediators: Drinking motives and alcohol cohol use (e.g., Ham, Stewart, Norton, & Hope, 2005; Zamboanga & outcome expectancies Ham, 2008; Zamboanga, Schwartz, Ham, Borsari, & Van Tyne, 2010). Ham (2009) found that social anxiety was positively related When examining the relationship between social anxiety and to positive social AOEs, which were positively related to hazardous alcohol use, it is important to consider variables through which alcohol use, and positive social AOEs partially mediated the rela- social anxiety may affect alcohol use variables (i.e., mediators), tionship between social anxiety and hazardous alcohol use. Other such as drinking motives (DMs) and alcohol outcome expectancies studies have also found that social anxiety has an indirect effect (AOEs). DMs, defined as the reasons why people choose to consume on ARPs through AOEs (Ham & Hope, 2005, 2006). alcohol (Grant, Stewart, & Mohr, 2009), have been conceptualized as the pathways through which more distal risk factors, such as 1.3. Potential moderators anxiety, depression, and sensation seeking, influence alcohol use at a given time (Cooper, 1994). Typically, four DMs are examined: The discrepant findings in the literature could be due to dif- enhancement of positive mood motives are internally motivated ferences in study characteristics. There are a number of possible and related to positive reinforcement (e.g., drinking to get a pleas- variables that could affect study results, perhaps the most obvious ant feeling or because a person finds it fun to drink); social rewards being participant characteristics. A recent study found that the motives are externally motivated and positively reinforcing (e.g., relationship between social anxiety and ARPs may be moderated drinking to be more sociable or to celebrate an event or occasion by gender (Norberg, Olivier, Alperstein, Zvolensky, & Norton, with others); conformity motives are externally motivated and neg- 2011; Norberg et al., 2009), so the proportion of the sample that atively reinforcing (e.g., drinking to fitinwithothersorduetopres- is female should be considered. Since no studies have examined sure from others); and coping with negative emotions motives are race or ethnicity as a potential moderator, it is important to consid- internally motivated and negatively reinforcing (e.g., drinking to er the possibility that the relationship between social anxiety and dealing with negative mood states such as depression or anxiety; alcohol-related variables may varybyrace.Additionally,theway Cooper, 1994). AOEs are defined as outcomes that an individual ex- participants were recruited and selected (e.g., if advertisements pects to occur as a result of consuming alcohol (Jones, Corbin, & were targeted to obtain certain individuals) could affect the char- Fromme, 2001). AOEs are learned through individuals' own alcohol acteristics of the sample. Ascertainment approach could result use as well as the effects that they observe and learn about from in a restricted range on a given variable (i.e., social anxiety or an others' experiences with alcohol (Jones et al., 2001). alcohol variable), but it is also possible that these variables moder- Studies with both non-college and college student samples have ate the relationship, as they may be stronger in certain subgroups shown that individuals high in social anxiety are more likely to (e.g., those high in social anxiety or those who drinking heavily and use alcohol as a way to reduce their anxiety in social situations, have been referred for problematic use). and they tend to avoid social situations where alcohol is not avail- The method of data collection is also important because in- able, compared to individuals with low social anxiety (Buckner person studies may result in a somewhat different sample than & Heimberg, 2010; Thomas, Randall, & Carrigan, 2003). DMs have on-line studies. For example, individuals high in social anxiety been shown to predict alcohol use and ARPs (e.g., Cooper, Frone, may try to avoid in-lab studies in order to avoid the required social Russell, & Mudar, 1995; Gonzalez, Bradizza, & Collins, 2009; Grant interaction. Additionally, because published studies often have larger et al., 2009; Ham & Hope, 2003; Park & Levenson, 2002). Using alco- effect sizes and are more likely to report significant effects than non- hol to cope with negative emotions appears to be particularly prob- published studies (Cooper, 2010; Lipsey & Wilson, 2001), it is important lematic and has been found to be correlated with suicidal ideation, to consider the possibility of publication bias when conducting a meta- depression, alcohol use, binge drinking, and ARPs (Gonzalez et al., analysis. 2009; Park & Levenson, 2002). Ham and Hope (2003) suggested Finally, characteristics of the measures used are important pos- that coping motives may be associated with drinking problems sible moderators. A variety of self-report questionnaire measures because individuals who use alcohol as a mechanism to cope with of social anxiety have been developed, and they differ in their rela- negative emotions lack safer, more effective, and longer-lasting tive focus on cognitive (e.g., fear of negative evaluation), behavioral coping mechanisms. In support of this hypothesis, students who (e.g., avoidance), and physiological symptoms. Furthermore, mea- drink to cope have been found to possess lower coping abilities sures differ on the types of social situations (e.g., performance and and higher positive alcohol expectancies (Park & Levenson, 2002). interaction) they include. Similarly, assessments of ARPs assess Therefore, social anxiety may increase the likelihood that individ- different types of consequences (e.g., social consequences, risk uals will drink alcohol for specific reasons, especially coping mo- behaviors, and personal consequences). Distinct aspects of social tives, which may in turn increase the likelihood of experiencing anxiety may be differentially related to alcohol-related variables. ARPs and, in turn, of developing AUDs in the future. In fact, coping For example, avoidance of social situations may be negatively cor- motives have been found to partially mediate the relationship related with many alcohol-related variables because a person may between social anxiety and both ARPs (Buckner & Heimberg, have few opportunities to drink alcohol socially and, in turn, experi- 2010; Norberg, Norton, Olivier, & Zvolensky, 2010)andsymptoms ence certain ARPs. In contrast, worry about negative evaluation of alcohol dependence (Ham et al., 2009). In another study, (without avoidance) may relate differently, as the person may use A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 2693 alcohol to reduce their anxiety in social situations and experience that effect sizes would be larger in published studies compared to ARPs as a result. Furthermore, one study found that the relationship unpublished studies. between social anxiety and ARPs differs by type of consequence (Norberg et al., 2011). Finally, the amount of time individuals 2. Materials and methods were asked to consider in their reporting could affect results. While shorter time frames are more accurate due to less demand 2.1. Operational definitions on memory, these shorter timeframes could be more affected by sporadic changes in alcohol use (e.g., attending a major sporting Social anxiety was defined as anxiety in, or avoidance of, social event) if they fall in the reporting timeframe. and performance situations, including anxiety related to evaluation by others in these situations (APA, 2000). Alcohol use was defined 1.4. Hypotheses as any measure of the consumption of beverages containing alcohol. Frequency of alcohol use was defined as the number of times a person In light of inconsistent findings in the extant literature on the consumed alcohol during a given period of time; whereas quantity of relationship between social anxiety and various alcohol-related alcohol use was defined as the number of standard alcoholic drinks variables in college students, we undertook a meta-analysis to consumed per drinking episode or per day. If the number of alcoholic clarify the nature of the relationship between social anxiety and drinks consumed over a period of time (e.g., a week or a month) was alcohol variables among college students. Specifically, we sought measured, this variable was considered a frequency by quantity index, to examine the relationship between social anxiety and four alcohol- because both number of drinks consumed per episode and frequency related variables: alcohol use, ARPs, DMs, and AOEs. Morris et al. of consumption affect this variable. Finally, the frequency of binge (2005) completed a qualitative review of the literature on social anxiety drinking was also included as an alcohol use variable. Binge drinking and alcohol use disorders and included separate sections on the was defined as five or more drinks in a single sitting for males and research in college students. However, a number of studies on the four or more drinks in a single sitting for females (per the National In- topic have been conducted since publication of that review. Because stitute on Alcohol Abuse and Alcoholism's Council Work Group's def- we found no previously published meta-analyses on this topic, we be- inition of binge drinking (NIAAA, 2003)). ARPs were defined as lieve this is the first meta-analysis to examine the relationship between problems or negative consequences that resulted from the consump- social anxiety and alcohol-related variables in college students. It is tion of alcohol. Finally, DMs were defined as reasons for consuming hoped that this meta-analysis will help to summarize the existing liter- alcohol, while AOEs were defined as expectations about the outcomes ature in order to help researchers plan future studies to further clarify of alcohol use. It is important to note that motives are distinct from this relationship. alcohol expectancies, as alcohol expectancies are the outcomes indi- Although alcohol may be used to decrease anxiety, it is impor- viduals anticipate will result from consuming alcohol regardless of tant to note that most college students consume alcohol in social whether an individual consumes alcohol in order to achieve these situations (e.g., parties), and one frequently used coping technique outcomes. for managing social anxiety is avoidance of situations that evoke fear and anxiety (APA, 2000). Therefore, we hypothesized that 2.2. Literature search social anxiety would be (1) positively correlated with the amount of alcohol consumed per drinking episode, but (2) negatively corre- Studies for this review were obtained through (a) computer lated with the frequency of alcohol use. Furthermore, we hypothe- searches of PsycNET and PubMed using the terms social anxiety, sized that (3) social anxiety would be positively related to ARPs, alcohol, and college students/undergraduates in any field through given the high rates of comorbidity between SAD and AUDs. June 10, 2012; (b) a manual search of the table of contents of Addic- Secondarily, we examined the relationships between social tion, Addictive Behaviors, Behavior Therapy, College Student Journal, anxiety and DMs and social anxiety and AOEs in order to examine Depression and Anxiety, Journal of American College Health, Journal potential mediators of the relationship between social anxiety and of Anxiety Disorders, Journal of College Student Development, Journal alcohol use or ARPs. It was expected that social anxiety would be of Studies on Alcohol and Drugs,andPsychology of Addictive Behaviors (4) positively related to coping and conformity motives for consum- in the 36 months prior to the study (i.e., from to June 2009 to the ing alcohol. It was expected that college students high in social anx- most recent issue available on-line as of June 10, 2012, including iety would use alcohol to cope with their anxiety in social situations, advanced on-line publication of articles that will be published where drinking is likely to occur. Additionally, because socially anx- in future issues, if available); (c) a manual search of the reference ious individuals are generally less assertive than non-socially lists of studies identified by the electronic and manual searches anxious peers (APA, 2000; Chambless, Hunter, & Jackson, 1982; that met inclusion criteria for the meta-analysis; (d) personal LeSure-Lester, 2001), they may be more likely to drink for conformi- communications with researchers who were the first author of at ty motives, as they may not be able to say no to alcohol when others least three reports that met inclusion criteria asking if they knew of pressure them to drink. Finally, it was hypothesized that (5) social additional studies or unpublished data; and (e) e-mail requests sent to anxiety would be positively related to both positive and negative Division 28 (Psychopharmacology and Substance Abuse) of the Ameri- alcohol expectancies. It was anticipated that individuals with social can Psychological Association and the Anxiety Disorders Special Interest anxiety would expect alcohol to have positive outcomes, such as in- Group (SIG) and the Addictive Behaviors SIG of the Association for Be- creased assertiveness or sociability. However, it was also expected havioral and Cognitive Therapies requesting relevant, unpublished that individuals with social anxiety would also report higher nega- studies. tive alcohol expectancies due to concerns about evaluation due to negative behaviors and, therefore, heightened concern about these 2.3. Study inclusion criteria and resulting pool of studies outcomes. Most moderator analyses were exploratory, without a priori In order to be included in the meta-analysis, a report had to (a) be hypotheses. However, given the results of Norberg et al. (2009, published in English or be available in English; (b) be empirically 2011), we expected the percentage of the sample that was female based (e.g., theoretical and review articles not including new data to be positively related to the effect size (i.e., when there were were not included); (c) only include college students, defined as stu- more females, the effect size would be larger), particularly for anal- dents enrolled in two-year or four-year undergraduate programs yses of ARPs. Due to concerns about publication bias, we hypothesized with a mean age of 18.0 to 24.0 years, as participants or have reported 2694 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706

219 total reports identified

3 reports excluded 10 reports excluded 25 reports excluded 112 reports excluded 18 reports excluded because they were because they were because sample was because social anxiety because the not available in not original not college sample/ and/or at least one correlation English empirical studies did not meet mean alcohol variable of coefficient was not age requirement interest not measured reported

50 total reports included

44 separate studies identified

23 studies included 23 studies included 19 studies included 30 studies included 10 studies included 15 studies included in SA and AU in SA and AU in SA and AU in SA and ARPs in SA and DMs in SA and AOEs quantity analyses frequency analyses frequency X analyses analyses (7 measured analyses (15 quantity analyses all 4 motives) measured positive AOEs and 9 measured negative AOEs)

Fig. 1. Flowchart illustrating the selection process and characteristics of included studies.

results separately for college students; (d) measure social anxiety almost perfect for five (50%) of the categorical variables. A table and at least one alcohol-related variable of interest (i.e., alcohol use, containing descriptions of each study is included in Appendix A. ARPs, DMs, or AOEs); and (e) report the correlation of social anxiety with the alcohol-related variable or provide sufficient information 2.4.1. Categorical moderators for the correlation to be computed. For studies identified by computer The categorical moderators were: recruitment/selection criteria searches, the abstract was reviewed for inclusion criteria; for the for the sample; data collection method (or type of assessment); manual searches of table of contents and reference lists, titles publication status; and the measures used to assess variables of were reviewed followed by an abstract review, if appropriate. Inclu- interest (e.g., social anxiety and ARPs). Studies were coded for sion criteria were evaluated sequentially (e.g., if the sample did not whether they recruited or selected participants high in social anx- include college students, the report was not included and other iety (e.g., treatment seeking sample, stating in recruitment mate- inclusion criteria were not evaluated). If all inclusion criteria rials that they were seeking individuals who feel anxious in social except (e) were met, we attempted to contact the corresponding situations), low in social anxiety, or high and low social anxiety author of the report to obtain the needed data. A study was defined (i.e., the researchers sought to exclude individuals with moderate as a set of data collected on a given sample. Because a single study levels of social anxiety; κ = .390). Because no studies recruited can span multiple publications or reports, multiple reports on a single or selected participants low in social anxiety, only two studies recruited study were coded together as a single study. or selected participants high in social anxiety, and five studies recruited A total of 219 reports (i.e., journal articles, doctoral dissertations, or selected participants high and low in social anxiety, this variable was etc.) were obtained using the search criteria described above. dichotomized for the purposes of analyses based on whether the study The first author, a doctoral student in a clinical psychology program, reviewed all identified reports for inclusion. Of these reports, 50 met the inclusion criteria, resulting in a sample of 44 separate Table 1 fl studies (see Fig. 1 for a ow chart of the inclusion and exclusion of Mean effect sizes for correlations between social anxiety and alcohol variables. studies in the meta-analysis). Alcohol variable Mean ES 95% CI Homogeneity analysis Q (df) 2.4. Study coding AU − Quantity −.0696⁎⁎⁎ −.0962 to −.0429 36.9347 (22)⁎ AU — Frequency −.1045⁎⁎⁎ −.1310 to −.0779 35.0271 (22)⁎ All 44 studies were coded by the first author, ten of which were AU — Frequency −.0777⁎⁎⁎ −.1122 to −.0430 2.3803 (5) randomly selected for reliability coding. A total of three raters were of Binge Drinking trained by the first author in the coding of the reports; two raters AU — Frequency × Quantity −.1081⁎⁎⁎ −.1473 to −.0686 40.7634 (18)⁎⁎ were undergraduate research assistants in psychology and one is a Index ARPs .0530⁎ .0126 to .0933 112.1600 (29)⁎⁎⁎ PhD level clinical psychologist. After reviewing the coding manual DMs — Coping .2363⁎⁎⁎ .1847 to .2866 23.1441 (9)⁎⁎ and joint coding of studies to establish initial consistency, the selected DMs — Conformity .2379⁎⁎⁎ .1746 to .2993 18.0091 (6)⁎⁎ studies were coded independently by each reliability coder. Discussion DMs — Enhancement .0318 −.0528 to .1160 29.5495 (6)⁎⁎⁎ — ⁎⁎⁎ to reach a consensus was used when there was a disagreement in DMs Social .0838 .0437 to .1236 7.0943 (6) AOEs — Positive .1457⁎⁎⁎ .1036 to .1872 33.5424 (14)⁎⁎ codes. For continuously coded items (e.g., sample size, percentages AOEs — Negative .1604⁎⁎⁎ .1091 to .2108 19.8318 (8)⁎ for gender and ethnicity), intraclass correlation coefficients (ICCs) were computed as a measure of reliability, and reliability for categorical Note. ⁎ p b 0.05. items was assessed using kappa. Using Landis and Koch's (1977) classi- ⁎⁎ p b 0.01. fication, kappa was fair for one (10%), substantial for four (40%), and ⁎⁎⁎ p b 0.001. A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 2695

Table 2 Tests of categorical moderators.

Alcohol Variable Variable and Class Between-Class Effect k Effect Size by Class Within-Class Effect

QB QW AU–Quantity Recruited/selected for social anxietya 2.8834 Not recruited/selected based on social anxiety 17 −.0778⁎⁎⁎ 11.1948 Recruited/selected based on social anxiety 6 −.0021 14.1457⁎ Recruited/selected for alcohol use/ARPs 6.3585⁎ Not recruited/selected based on AU/ARPs 6 −.0938⁎⁎⁎ 1.6480 Non-drinkers excluded 11 −.0708⁎⁎ 8.9424 Recruited/selected for high AU/ARPs 6 .0174 12.2348⁎ Data Collection Methodb 1.8489 In person 11 −.0371 20.3505⁎ Other 10 −.0807⁎⁎⁎ 3.4809 Published/Peer-Reviewedc 1.5328 No 8 −.0480⁎ 2.0656 Yes 15 −.0823⁎⁎⁎ 24.1458⁎ Social Anxiety Measure .1192 Social Interaction Anxiety Scale (SIAS) 7 −.0698⁎⁎ 5.0867 Other 7 −.0733⁎⁎ 18.1344⁎⁎ Multiple 9 −.0612⁎ 1.7321 AU – Frequency Recruited/selected for social anxiety 1.2455 Not recruited/selected based on social anxiety 17 −.1098⁎⁎⁎ 13.0455 Recruited/selected based on social anxiety 6 −.0616 13.1141⁎ Recruited/selected for alcohol use/ARPs 7.6226⁎ Not recruited/selected based on AU/ARPs 8 −.0941⁎⁎⁎ 8.6160 Non-drinkers excluded 11 −.1281⁎⁎⁎ 4.4075 Recruited/selected for high AU/ARPs 4 .0630 8.8322⁎ Data Collection Method 7.3927⁎⁎ In person 11 −.0437 17.1990 Other 9 −.1252⁎⁎⁎ 3.0719 Published/Peer-Reviewed .6380 No 8 −.1183⁎⁎⁎ 5.6678 Yes 15 −.0950⁎⁎⁎ 20.0630 Social Anxiety Measure .9887 Social Interaction Anxiety Scale (SIAS) 8 −.1132⁎⁎⁎ 1.8228 Other 7 −.0794⁎⁎ 16.5456⁎ Multiple 8 −.1110⁎⁎⁎ 5.8186 AU – Frequency X Quantity Index Recruited/selected for social anxiety 4.8010⁎ Not recruited/selected based on social anxiety 16 −.1191⁎⁎⁎ 14.1515 Recruited/selected based on social anxiety 3 .0633 6.7166⁎ Recruited/selected for alcohol use/ARPs .5473 Not recruited/selected based on AU/ARPs 8 −.1044⁎⁎ 5.3660 Non-drinkers excluded 3 −.1257⁎⁎ .5188 Recruited/selected for high AU/ARPs 7 −.0780 14.7582⁎ Data Collection Method .9872 In person 12 −.1224⁎⁎⁎ 19.3049 Other 6 −.0795⁎ 2.6431 Published/Peer-Reviewed .0033 No 6 −.1094⁎⁎ 3.7576 Yes 13 −.1068⁎⁎⁎ 19.3020 Social Anxiety Measured 2.8373 Social Interaction Anxiety Scale (SIAS) 3 −.0465 1.2691 Other 7 −.1369⁎⁎⁎ 16.2949⁎ Multiple 9 −.1150⁎⁎⁎ 4.1272 ARPs Recruited/selected for social anxiety 5.5085⁎ Not recruited/selected based on social anxiety 25 .0355 31.4357 Recruited/selected based on social anxiety 5 .1773⁎⁎ 2.5226 Recruited/selected for alcohol use/ARPse 3.0330 Not recruited/selected based on AU/ARPs 14 .0281 14.2537 Non-drinkers excluded 10 .0712⁎ 7.7173 Recruited/selected for high AU/ARPs 5 .1386⁎ 8.4025 Data Collection Method 1.1586 In person 17 .0352 19.4128 Other 11 .0840⁎ 10.7396 Published/Peer-Reviewed 4.5616⁎ No 10 −.0079 12.8096 Yes 20 .0819⁎⁎⁎ 20.0950 ARP Measure .8850 Rutgers Alcohol Problems Index (RAPI) 11 .0692 16.7837 Alcohol Use Disorders Identiﬁcation Test (AUDIT) 4 .0863 1.8331 Other 15 .0356 13.1513 Social Anxiety Measuref 5.8295 Social Interaction Anxiety Scale (SIAS) 8 .1197⁎⁎ 5.6870 Other 9 −.0044 7.2939 Multiple 13 .0526 18.3437

(continued on next page) 2696 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706

Table 2 (continued) Alcohol Variable Variable and Class Between-Class Effect k Effect Size by Class Within-Class Effect

QB QW DMs – Coping Recruited/selected for social anxiety 8.4138⁎⁎ Not recruited/selected based on social anxiety 8 .2142⁎⁎⁎ 7.7337 Recruited/selected based on social anxiety 2 .4101⁎⁎⁎ .0196 Data Collection Method 1.2424 In person 4 .2047⁎⁎⁎ 6.0298 Other 5 .2693⁎⁎⁎ 3.4113 Published/Peer-Reviewed .0481 No 3 .2285⁎⁎⁎ .3371 Yes 7 .2416⁎⁎⁎ 9.9422 Social Anxiety Measureg 4.7139 Social Interaction Anxiety Scale (SIAS) 3 .2603⁎⁎⁎ 5.6721 Other 3 .1627⁎⁎⁎ .7733 Multiple 4 .2799⁎⁎⁎ 1.9890 AOEs – Positive Recruited/selected for social anxiety 6.3901⁎ Not recruited/selected based on social anxiety 13 .1288⁎⁎⁎ 14.2559 Recruited/select based on social anxiety 2 .2845⁎⁎⁎ .2211 Data Collection Method .4548 In person 7 .1260⁎⁎⁎ 5.7786 Other 5 .1609⁎⁎⁎ 5.8274 Published/Peer-Reviewed 1.4684 No 3 .0965⁎ .5770 Yes 12 .1589⁎⁎⁎ 14.8239 AOEs measure 4.6290 Comprehensive Effects of Alcohol Scale (CEOA) 4 .0668 3.1487 Other 5 .1908⁎⁎⁎ 4.8800 Multiple 6 .1548⁎⁎⁎ 3.1724

Note. ⁎ p b 0.05. ⁎⁎ p b 0.01. ⁎⁎⁎ p b 0.001. a Q(1) = 4.4417, p = .0351; Did not recruit/select based on social anxiety: mean ES = −.0890, p b .001, Qw(16) = 17.0144, p = .3847; Recruited/selected based on social anxiety: mean ES = −.0035, p = .9297, Qw(5) = 15.4785, p = .0085. b Q(1) = 4.6941, p = .0303; In-person: mean ES = −.0411, p = .0641, Qw(10) = 24.7053, p = .0059; Other: mean ES = −.0927, p b .001, Qw(9) = 7.0540, p = .6315. c Q(1) = 4.7407, p = .0295; Not published/peer-reviewed: mean ES = −.0510, p = .0042, Qw(7) = 2.8875, p = .8952; Published/peer-reviewed: mean ES = −.0939, p b .001, Qw(14) = 29.3064, p = .0095. d Q(2) = 7.5034, p = .0235; SIAS: mean ES = −.0536, p = .0364, Qw(2) = 3.2406, p = .1978; Other: mean ES = −.1364, p b .001, Qw(6) = 22.5319, p = .0010; Multiple: mean ES = −.1276, p b .001, Qw(8) = 7.4874, p = .4851. e Q(2) = 8.5224, p = .0141; Did not recruit/select based on alcohol use/ARPs: mean ES = .0077, p = .5945, Qw(13) = 54.3965, p b .001; Non-drinkers excluded: mean ES =

.0474, p = .0023, Qw(9) = 25.8041, p = .0022; Recruited/selected for high alcohol use/ARPs: mean ES = .1066, p = .0020, Qw(4) = 20.1040, p b .001. f Q(2) = 16.8759, p b .001; SIAS: mean ES = .0936, p b .001, Qw(7) = 19.5073, p = .0067; Other: mean ES = −.0126, p = .4641, Qw(8) = 26.3988, p b .001; Multiple: mean

ES = .0248, p = .1139, Qw(12) = 49.3781, p b .001. g Q(2) = 9.1218, p = .0105; SIAS: mean ES = .2606, p b .001, Qw(2) = 9.5358, p = .0085; Other: mean ES = .1689, p b .001, Qw(2) = 1.4695, p = .4796; Multiple: mean

ES = .2704, p b .001, Qw(3) = 3.0170, p = .3890.

recruited or selected based on social anxiety. Studies were also coded for measures for the purposes of analyses. The Rutgers Alcohol whether they excluded non-drinkers, recruited or selected participants Problems Index (RAPI; White & Labouvie, 1989) and Alcohol high in alcohol use/ARPs (e.g., students referred to campus offices for Use Disorders Identification Test (AUDIT; Babor, Higgins-Biddle, campus alcohol violations), recruited or selected participants low in Saunders, & Monteiro, 2001) were the most commonly used mea- alcohol use/ARPs (e.g., excluding students referred to campus offices sures of ARPs (used in 27.3% and 18.2% of studies, respectively), for campus alcohol violations), or recruited or selected participants and this variable was collapsed to RAPI, AUDIT, or other for the high and low alcohol use/ARPs (κ = .674). Because only one study purposes of analyses. The Drinking Motives Questionnaire or Drink- recruited or selected participants low in alcohol use/ARPs (Ham, ing Motives Questionnaire — Revised (DMQ-R; Cooper, 1994)were 2003; Ham & Hope, 2006), this study was excluded from the moderator the most commonly used measures of DMs, and only two studies analyses for recruitment/selection based on alcohol use/ARPs. reporting motives did not use the DMQ or DMQ-R. Therefore, The data collection method was coded as occurring in-person this variable was not analyzed as a potential moderator. Finally, the (e.g., in a lab or classroom setting) or not in-person (e.g., internet Comprehensive Effects of Alcohol Scale (CEOA; Fromme, Stroot, & surveys and mail questionnaires). If any portion of the relevant Kaplan, 1993) was the most commonly used single measure of AOEs data collection was in-person, then the data collection method (used in 11.4% of studies); this variable was collapsed into CEOA, was coded as “in-person” (κ = .747). Whether or not any portion other measure, and multiple measures for the purposes of analyses. of the study was published in a peer-reviewed source (i.e., not published, published–not peer-reviewed, or published–peer-reviewed) 2.4.2. Continuous moderators was coded (κ = .726). All included studies were either not published Continuous moderators assessed were gender, race/ethnicity, or were published in a peer-reviewed journal. and reporting timeframe. The percentage of the sample that was The self-report measures of social anxiety (κ = .901), ARPs female was coded for each study (ICC = 1.000). The race and ethnic- (κ = 1.000), DMs (κ = 1.000), and AOEs (κ = 1.000) were coded. ity of the study samples were coded as percentages of the sample The Social Interaction Anxiety Scale (SIAS; Mattick & Clarke, 1998) that identified as various racial/ethnic groups. However, because was the most commonly used single measure of social anxiety which racial/ethnic categories were reported across studies varied, (used in 29.5% of studies). Due to the wide range of other measures only the percentage of the sample that was Caucasian/European was used, this code was collapsed to SIAS, other measure, or multiple used in analyses (ICC for percentage Caucasian/European = .754). A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 2697

The reporting timeframe for alcohol use (κ = 1.000) and ARPs (κ = exceeds what would be expected from sampling error alone (Lipsey .728) was coded in days (however, because most studies used standard & Wilson, 2001). The significance of individual predictors is evaluated reporting timeframes, rater agreement was assessed with kappa). by examining the significance of the regression coefficients. Reporting timeframe was defined as the period of time participants Due to the significant heterogeneity in effect sizes for almost were asked to think about when completing the measures (e.g., if a all relationships tested, mixed effects models were used to test the ef- participant was told to report the average number of drinks consumed fects of moderator variables (Cooper, 2010; Lipsey & Wilson, 2001). each day of the week over the past month, this was coded as 30 days). However, because of the small number of studies that met inclusion criteria and the fact that mixed effects models have lower statistical 2.5. Effect sizes power (i.e., are more stringent) to detect the effects of moderator variables (Lipsey & Wilson, 2001), when a fixed effects model Pearson zero-order correlations were used as the effect size indi- resulted in a significant effect and the mixed effects model did not, cators. If the correlation coefficientwasnotincludedinthereport we have noted these effects with a footnote in the table. but sufficient information was included to estimate the correlation coefficient, the Practical Meta-Analysis Effect Size Calculator was 3. Results used to compute the estimate (Lipsey & Wilson, 2001). If correlation coefficients were reported separately for males and females, then 3.1. Mean effect sizes and homogeneity analyses the correlations were transformed using Fisher's Zr-transformation (Hedges & Olkin, 1985), weighted by their inverse variance weight Mean effect sizes and results of the homogeneity analyses are (Lipsey & Wilson, 2001), averaged, and then transformed back to a presented in Table 1. Social anxiety was significantly negatively cor- correlation to obtain the overall correlation between the two vari- related with all measures of alcohol use (i.e., quantity, frequency, ables for the sample. If the study used multiple measures of social anx- frequency of binge drinking, and frequency by quantity indices). In iety, the average correlation for each alcohol variable was computed contrast, the mean correlation between social anxiety and ARPs using the same procedure described above. Average effect sizes were was significant and positive. Additionally, social anxiety was signif- computed in these cases because effect sizes in a meta-analysis are icantly, positively correlated with three of the four DMs (i.e., coping, assumed to be independent. Therefore, when multiple effect sizes are conformity, and social motives), but it was not significantly cor- reported for a single relationship, an average effect size can be computed related with enhancement motives. Finally, the mean correlation to prevent inclusion of dependent effect sizes in the meta-analysis between social anxiety and both positive and negative AOEs was (Cooper, 2010). If a study reported alcohol use or ARPs for multiple significant and positive. timeframes, only the correlation for the longest timeframe was coded The homogeneity analyses for the majority of the effect sizes were in order to allow for assessment of the effect of reporting timeframe in significant, indicating that there is a significant amount of heteroge- moderator analyses. neity and the amount of variability exceeds what would be expected simply by sampling error (Lipsey & Wilson, 2001). However, the 2.6. Statistical analyses homogeneity test was not significant for the relationships between social anxiety and frequency of binge drinking and social anxiety and All data were entered into Predictive Analytics Software Statistics social motives. 18.0 (PASW Statistics 18.0) for analysis, and macros designed for meta-analyses (Wilson, 2010)wereusedtoconducttheanalyses. 3.2. Analyses of moderator variables Before being analyzed, correlations were transformed using Fisher's

Zr-transformation (Hedges & Olkin, 1985) and weighted by their in- Moderator effects were only examined for those effect sizes reported verse variance weight (i.e., n − 3; Lipsey & Wilson, 2001). The in at least ten studies, due to concerns about power to find a significant MeanES macro (Wilson, 2010) was used to compute the mean effect effect, if present, and concerns about having a sufficient number of size for each relationship, the p-value for the mean effect size, and studies per group when categorical variables were examined. Therefore, the 95% confidence interval of the mean effect size. The random ef- moderator variables were not examined for the relationships between fects model estimates were computed. Homogeneity analyses were social anxiety and frequency of binge drinking, conformity motives, also conducted using this macro in order to determine whether the enhancement motives, social motives, or negative expectancies. Tests of variance in effect sizes for each relationship is statistically different categorical moderators could only be run when there were at least two from what would be expected if only sampling error accounted for studies per group; therefore, some moderators could not be tested for variability in the effect sizes (Cooper, 2010). some outcomes. Results of the tests for categorical moderators are The effects of categorical moderator variables were tested using presented in Table 2, and the results of the tests for continuous modera- the MetaF macro (Wilson, 2010), an analog to an ANOVA (Lipsey tors are presented in Table 3. All potential moderators that were explored & Wilson, 2001). This macro computes the between-class effects are included in the respective tables.

(QB), which tests the homogeneity of studies between groups, and When the correlation between social anxiety and alcohol use the within-class variability (QW), which tests the homogeneity within quantity was examined, only recruitment/selection based on alcohol the classes. If the between-class effect is significant (i.e., homogeneity use and/or ARPs was a significant moderator. While studies that did between classes is rejected), and the within-class variability is not not recruit or select participants based on alcohol use and/or ARPs significant, then the moderator adequately explains the variability and studies that only excluded non-drinkers had a significant nega- observed in the effect sizes; however, if the within-class variability tive mean correlation, studies that recruited or selected participants is significant, the moderator explains some of the variability, but high in alcohol use and/or ARPs had a non-significant mean correla- there is still significant variability remaining that was not explained tion, suggesting that the relationship was stronger for studies that by that particular moderator (Lipsey & Wilson, 2001). The effects did not recruit or select participants for elevated levels of alcohol of continuous moderator variables were tested using the MetaReg use and problems. macro (Wilson, 2010), which runs a weighted regression analysis. Recruitment/selection based on alcohol use/ARPs and data collec- In these analyses, the homogeneity test for the regression model tion method were significant moderators of the size of the correlation

(QR) evaluates whether the set of predictors explains a signiﬁcant between social anxiety and alcohol use frequency. As with the proportion of the variability in effect sizes, and the homogeneity analyses of alcohol use quantity, studies that recruited participants test of the residual (QE) evaluates whether the unexplained variance high in alcohol use/ARPs did not have a signiﬁcant mean correlation, 2698 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706

Table 3 Tests of continuous moderators.

Alcohol variable Predictors Regression model homogeneity (QR) Residual homogeneity (QE) β AU — Quantity 4.1792 20.8590 Alcohol use reporting timeframe .3133 % of sample female −.1623 % of sample Caucasian .2251 AU — Frequency 5.8268 14.7150 Alcohol use reporting timeframe .3018 % of sample female −.6047⁎ % of sample Caucasian .0219 AU — Frequency × Quantity Index 1.3893 10.6829 Alcohol use reporting timeframe −.1219 % of sample female −.3995 % of sample Caucasian −.2059 ARPs 6.7195 16.0180 ARPs reporting timeframe −.4665⁎ % of sample female a .2802 % of sample Caucasian .0818 DMs — Coping 1.3620 8.8857 % of sample female .1682 % of sample Caucasian −.3423 AOEs — Positive .0084 10.7794 % of sample female −.0149 % of sample Caucasian .0171

Note. ⁎ p b 0.05. a Q(3) = 22.7845, p b .001; β = .3747, p = .0034.

while studies that did not recruit or select participants based on on social anxiety was significant. While both types of studies had signif- alcohol variables or that excluded non-drinkers had significant icant positive mean effect sizes, studies that recruited or selected partic- negative mean correlations. Studies that did not require participants ipants based on social anxiety had a larger mean effect size. to complete measures in-person had a significant negative mean Finally, when the relationship between social anxiety and positive effect, but studies that did require in-person participation did AOEs was examined, only recruitment/selection based on social anxiety not have a significant mean effect. Additionally, percentage of was a significant moderator. Similar to the results of the analyses with thesamplethatwasfemalewasasignificant moderator, even coping motives, both types of studies resulted in a significant positive though the regression model did not explain a significant amount mean effect size, but the mean effect size for studies that did recruit ofthevarianceineffectsizes.Theeffectsizewaslargerinstudies or select participants based on level of social anxiety was larger. that included a larger percentage of female participants. It is important to note that the relationship between social anxiety 4. Discussion and frequency of alcohol use was negative. As such, studies using samples with more females found a stronger, inverse relationship This study examined the relationship between social anxiety betweensocialanxietyandfrequencyofalcoholuse. and alcohol variables in college students. The mean effect size of the When the correlation between social anxiety and frequency relationship between social anxiety and all measures of alcohol use by quantity indices of alcohol use was examined, only recruit- (i.e., quantity, frequency, frequency of binge drinking, and frequency ment or selection based on social anxiety was a significant by quantity indices) was significant and negative. These findings moderator. Studies that did not recruit or select participants indicate that socially anxious college students consume less in a typ- based on level of social anxiety had a significant negative mean ical drinking episode, drink alcohol less frequently, binge drink less effect size, while those that did resulted in a non-significant frequently, and consume fewer drinks over periods of time (e.g., a mean correlation. typical week or month) than their less socially anxious peers. Recruitment/selection based on social anxiety and publication The mean effect size for alcohol use quantity was in the opposite status were significant moderators of the effect size for the relation- direction as hypothesized. It may be that individuals who are socially ship between social anxiety and ARPs. Studies that did recruit or anxious consume only enough alcohol to reduce their anxiety; they select participants based on their level of social anxiety had a sig- may intentionally not drink in excess due to concerns about excessive nificant positive mean correlation, while those studies that did not alcohol use negatively affecting their social performance. Even if had a non-significant mean correlation. Additionally, published socially anxious individuals do not consume more drinks per episode, studies had a mean effect size that was significant and positive, they may still drink to reduce their anxiety. Therefore, for college while unpublished studies did not have a significant mean effect students who are high in social anxiety, number of alcoholic drinks size. Although the overall regression model was not significant, consumed may not be elevated, particularly compared to college reporting timeframe for ARPs was a significant moderator of effect peers, since heavy alcohol use is common among college students size, with a negative relationship between reporting timeframe and (e.g., Benton et al., 2004; Johnston et al., 2010). effect size. The finding that social anxiety was negatively related to frequency When moderators of the relationship between social anxiety of alcohol use supported our hypothesis. The importance of avoidance and coping motives were examined, only recruitment/selection based as a coping strategy must be considered in the relationship between A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 2699 social anxiety and alcohol use. A vast majority of drinking situations, participants based on social anxiety was a significant moderator particularly for college students, are in social settings (e.g., parties, of effect size in studies of the relationship between social anxiety meeting friends at a bar, and tailgating). Individuals high in and frequency by quantity indices, ARPs, coping motives, and posi- social anxiety may choose to avoid these situations the majority of tive AOEs. For studies that examined ARPs, coping motives, and the time and only use alcohol as a coping technique when they ei- positive AOEs, when participants were recruited or selected based ther choose to be in the situation or feel they cannot avoid the on social anxiety, the effect size was larger. Many of the studies situation. that recruited or selected participants based on social anxiety It is interesting to note that social anxiety had a small, but sig- sought a group of participants high in social anxiety and a group nificant, positive mean correlation with ARPs, despite being nega- of participants low in social anxiety; methodologically, creating ex- tively related to alcohol use. This finding is important, as AUDs treme groups may have enhanced their ability to detect an effect. are defined not only by alcohol use but primarily by problems In contrast, when the outcome was the relationship between social resulting from use (APA, 2000; Buckner et al., 2006). Therefore, anxiety and frequency by quantity indices of alcohol use, a larger this positive correlation is consistent with the prior research negative effect was found when studies did not recruit or select suggesting high rates of comorbidity of SAD and AUDs (see Morris participants based on social anxiety, suggesting that having a full et al. (2005),forareview). range of social anxiety in the sample enhances the ability to find Consistent with our hypotheses, the mean correlations between an effect. social anxiety and coping motives and social anxiety and conformi- Consistent with the finding that recruiting or selecting partic- ty motives were significant and positive. College students high in ipantsbasedonsocialanxietydecreasedtheoveralleffectsize social anxiety were expected to endorse more coping motives be- between social anxiety and frequency by quantity indices of cause it was expected that they would drink to reduce their anxiety alcohol use, data collection method was a significant moderator in social situations. Furthermore, individuals with SAD have higher of the relationship between social anxiety and frequency of rates of other anxiety disorders and mood disorders than individ- alcohol use. Studies that did not require in-person participation uals without SAD (Grant et al., 2005), and social anxiety in college resulted in a significant negative mean effect, while in-person students is positively associated with self-reported depression and studies did not have a significant mean effect. It is possible general worry (Schry, Roberson-Nay, & White, 2012). The high that socially anxious individuals avoid in-person studies due to rates of comorbidity may further result in more endorsement fears of interacting with the researcher, which could restrict of coping motives, as individuals with SAD may also be more likely the range of social anxiety in the sample and decrease power to consume alcohol to cope with multiple types of negative affect to find an effect. (e.g., related to co-occurring depression). The positive correlation Recruitment or selection based on alcohol use or ARPs was a sig- with conformity motives is consistent with the notion that individ- nificant moderator in studies that examined the relationship between uals with SAD struggle to be assertive (APA, 2000; Chambless social anxiety and quantity of alcohol use and frequency of alcohol et al., 1982; LeSure-Lester, 2001). Since alcohol use, and even use. In both cases, when studies recruited or selected a sample that heavy alcohol use, is arguably normative among college students was high in alcohol use or ARPs, the mean effect size was not signifi- (Johnston et al., 2010), it is likely that students in situations cant, but studies that either did not recruit or select in this way or that where alcohol is consumed perceive social pressure to consume only excluded individuals that did not drink resulted in a significant alcohol. Socially anxious individuals may be more likely to give in negative effect. These results indicate that when only individuals to this social pressure due to their poor assertion skills. Although with possible problematic use are included, social anxiety is not related not predicted, when examining the content of questions that assess to alcohol use. social motives, it is not surprising that there was a significant and When the relationship between social anxiety and ARPs was positive mean correlation with social anxiety, as these items ask examined, whether or not the study was published significantly about drinking alcohol to make social gatherings more fun and moderated the relationship between social anxiety and ARPs, being more sociable (Cooper, 1994). For individuals who are high with published studies having a significant positive mean effect in social anxiety, reducing their social anxiety may be related to size and unpublished studies having a non-significant mean ef- making social situations more enjoyable and being more sociable fect size. Therefore, concerns about publication bias are only (i.e., if they feel less anxious and inhibited, they can be more present for studies of social anxiety and ARPs. sociable). In most analyses, the continuous moderators examined did Finally, as hypothesized, social anxiety was positively related to not have a significant effect. However, the timeframe used for both positive and negative AOEs. Although intuitively contradictory, reporting of ARPs was a significant moderator of the relationship the finding that social anxiety was significantly and positively related between social anxiety and ARPs. It is interesting to note that the to both types of AOEs is consistent with prior research that many regression coefficient was negative, indicating that as the reporting types of AOEs are positively correlated, including positive and neg- timeframe increased, the effect size decreased. It is possible that ative AOEs (e.g., Ham & Hope, 2006; Ham et al., 2005). Given that longer reporting timeframes dilutes the effect because there is sig- social anxiety is positively associated with both coping and social nificant variability in college student drinking (e.g., students may motives, it is not surprising that social anxiety is positively related drink less when they go home on extended breaks), and longer to positive AOEs, as it may be argued that in order to be motivated timeframes may be more likely to include these times. It is also pos- to use alcohol for a certain reason, a person must expect that sible that college students struggle to recall drinking information, alcohol use will result in the positive, desired outcome. Consistent particularly ARPs for extended periods, since college students ex- with concerns about negative evaluation as a defining feature perience high rates of problems (Mallett et al., 2011); this difficulty of SAD (APA, 2000), we also found a significant positive relation- may lead to less accurate reporting for extended time periods. The ship with negative AOEs. It is possible that these negative AOEs percentage of the sample that was female was only a moderator of may mediate the negative relationship between social anxiety the relationship between social anxiety and frequency of alcohol and alcohol use quantity, as socially anxious individuals may use, with the effect size being larger (i.e., more negative) when drink less due to concerns about negative outcomes, particularly the percentage of females was greater. The fact that males tend social outcomes. to drink more than females (Ham & Hope, 2003) may explain this Multiple moderator variables had significant effects on the finding, as males may be more likely to see alcohol use as norma- observed effect sizes. Whether a study recruited or selected tive and therefore socially anxious male undergraduates may be 2700 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 more likely to go out drinking, while socially anxious female under- Researchersshouldalsoattempttodeterminewhysocial graduates may avoid these situations more. anxiety is positively related to ARPs despite being negatively related to alcohol use. If individuals high in social anxiety are not drinking more than less socially anxious peers, it is important 4.1. Limitations that we determine what about social anxiety makes individuals more vulnerable to ARPs. One important step is to determine Only 44 studies were included, and the number of studies in whether social anxiety is positively related to all types of ARPs, any given analysis was even smaller, ranging from six (for frequency or if it is differentially related to ARPs based on type (e.g., subjec- of binge drinking) to 30 (for ARPs). However, Lipsey and Wilson tively perceived problems such as doing something embarrassing). (2001) indicate that meaningful meta-analyses can be conducted with as few as two or three studies. As a result, it is important to keep in mind that although many of the moderators examined were 4.3. Clinical implications not significant, we may not have had sufficient power to detect significant effects, particularly in mixed effects models, which have lower It is important for clinicians who work with college students to power than fixed effects models. Additionally, the fact that measures note that although social anxiety is negatively associated with mea- of social anxiety had to be trichotomized into one group of studies sures of alcohol use, it appears to be positively associated with that used the most commonly used measure of social anxiety ARPs. Therefore, clinicians should consider asking about ARPs regard- (i.e., SIAS; Mattick & Clarke, 1998), a group of studies that used a less of an individual's level of alcohol consumption because they may fi single other measure, and a group of studies that used multiple mea- be experiencing signi cant problems related to alcohol use, even in sures was a significant limitation. It is likely that different measures of the absence of heavy use, particularly if they have SAD. Social anxiety fi social anxiety result in different effect sizes (e.g., measures that focus is also signi cantly related to drinking motives. Coping motives ap- on behavioral avoidance, as opposed to fear of negative evaluation, pear to particularly problematic (e.g., Cooper et al., 1995; Gonzalez may result in stronger negative correlations for alcohol use frequen- et al., 2009; Park & Levenson, 2002) and have been shown to mediate cy), but the number of different measures used and the low frequency the relationship between social anxiety and ARPs (Buckner & with which some measures were used did not allow for this type Heimberg, 2010; Norberg et al., 2010). It is likely that by helping indi- of analysis. Using the percentage of the sample that was female to viduals to develop other coping techniques and reducing their coping examine the moderating effect of gender poses a third limitation. motives for alcohol use, clinicians could help individuals high in social It would have been ideal to examine the correlation coefficients anxiety to reduce the number of ARPs they experience or prevent separately for males and females, but too few studies reported the them from experiencing ARPs. effect sizes separately for each gender. The percentage of females in the sample may not have affected the overall correlations sufficiently 4.4. Conclusions for an effect to be detected, even if the effects do differ for the two genders. This quantitative review examined the relationship between social anxiety and alcohol variables in college students. While many of the average effect sizes are small, the relationship between social 4.2. Recommendations for future research anxiety and alcohol variables is still important both clinically and sci- entifically because of the pattern of the results. Specifically, despite Researchers should consider the role of social avoidance when the fact that social anxiety was negatively related to quantity and fre- examining the relationship between social anxiety and alcohol var- quency of alcohol use, quantity by frequency measures of alcohol use, iables. There is a social component to the use of many substances, and frequency of binge drinking, it was positively related to ARPs. and this may be particularly true for alcohol use in college students. Therefore, research on this phenomenon is necessary to identify Failure to consider behavioral avoidance of social situations may interventions to reduce the increased number of ARPs experienced lead to misleading conclusions about the relationship between by those high in social anxiety, since their level of alcohol use is not social anxiety and alcohol use. For example, it may be that social elevated. Social anxiety was also positively related to coping, con- anxiety is negatively related to frequency of alcohol use, but social- formity, and social motives for alcohol use, which may put socially ly anxious individuals may be more likely to drink alcohol when anxious college students at risk for problematic use. There was also they are in social situations or prior to social events. Therefore, a positive relationship with positive and negative outcome expectan- it may be important to examine the percentage of times an individ- cies. Both DMs and AOEs are important variables to continue to exam- ual consumes alcohol out of the total number of times he/she ine in research as possible mediators of the relationship between was in a situation where others were drinking. It may also be social anxiety and alcohol variables. important to examine the relationship between social anxiety and drinking prior to social events, which may be related to ARPs (Nitka & O'Connor, 2011). It may also be important to consider Role of funding sources This study was not funded. other situations in which socially anxious individuals may be likely to drink prior to the event (e.g., prior to presentations or tests), Contributors which may be related to ARPs even if alcohol consumption is not Amie R. Schry was responsible for writing the coding manual, conducting the liter- excessive. ature search, coding all studies, conducting the analyses, and writing the first draft of It is also recommended that studies should be powered to the manuscript. Susan W. White consulted on the design and coding protocol, served examine the potential moderating effects of gender. In addition to as a reliability coder, and assisted in writing the manuscript. the fact that males and females differ on alcohol variables (Ham & Hope, 2003), recent research suggests that the relationship Conflict of interest fl between social anxiety and substance use problems may vary The authors declare that they have no con icts of interest. by gender (e.g., Buckner, Heimberg, & Schmidt, 2011; Buckner & Turner, 2009; Norberg et al., 2009). While some studies con- Acknowledgments We would like to thank Danny Axsom, PhD, Bethany C. Bray, PhD, and trol for gender, if the effects are different, simply entering gen- Robert Stephens, PhD, for consultation on the planning and design of this study. der as a covariate will not allow the researcher to detect these We would also like to thank Elmira Hamidi and Caitlyn Berry for serving as reliability differences. coders. Appendix A Characteristics of studies included in the meta-analysis.

Report information Sample size Rec./sel. % Female % Caucasian Data coll. Pub./ SA meas. AU ARP meas. ARP Motive AOEs meas. Correlations method peer-review time-frame time-frame meas. report

Bagner, Storch, 215 None 76% 72% In person Yes SAS-A n/a PAI Not rep. n/a n/a ARP = .09 and Preston (2007) Bella and 413 None 34% Not rep. In person Yes WMH-CIDI Not reported n/a Not rep. n/a n/a AU Freq. = .05 Omigbodun (2009) Buckner, Ecker, 172 High/low SA 76% 87% Other Yes SIAS Month RAPI Month n/a n/a AU Quant. = −.07 et al. (2011) Non-drinkers AU Freq. = −.05 excluded ARP = .26 Buckner and 120 High/low SA; 81% 85% Other Yes SIAS Month RAPI Month DCSAS n/a AU Quant. = −.14 Heimberg (2010) Non-drinkers AU Freq. = −.15 excluded ARP = .21 Cope = .42 Buckner and 232 High/low SA Not rep. Not rep. Other Yes SIAS Month RAPI Month n/a SIDS AU Freq. = −.16

Matthews (2012) ARP = .18 2690 (2013) 38 Behaviors Addictive / White S.W. Schry, A.R. Pos. = .31 Neg. = .31 Buckner et al. 293 Non-drinkers 58% 86% In person Yes SIAS Month RAPI Month DMQ n/a AU Quant. = −.08 (2006) excluded AU Freq. = −.06 ARP = .17 Cope = .13 Enh. = .18 Soc. = .13 Burke 73 High alcohol use/ 59% Not rep. In person Yes IAS Month n/a n/a n/a n/a AU Freq. = −.18 and Stephens (1997) problems AU FQ = −.13 Clerkin 730 Non-drinkers 60% 66% Not rep. Yes SAD Month YAAPST Year DMQ n/a AU Quant. = −.11 and Barnett (2012) excluded AU Freq. = −.16 AU FQ = −.16 ARP = −.04 Cope = .18 Conf. = .19 Enh. = −.11 Soc. = .02 Dhuse (2005) 334 None 69% Not rep. In person No SIAS and SAD 2 weeks YAAPST Year n/a AEQ AU FQ = −.08

− –

and EDA ARP = .03 2706 Pos. = .15 Neg. = .17 Eggleston, 284 Non-drinkers 58% 86% Not rep. Yes SIAS 1 week RAPI Month n/a CEOA AU Freq. = −.12 Woolaway-Bickel, excluded and AESES AU Binge = −.12 and Schmidt (2004) ARP = .11 Pos. = .22 Neg. = .19 Gilles (2005), 118 None 61% 86% Other Yes SIAS/SPS total Not rep. AUDIT Not rep. n/a CEOA AU FQ = −.04 Gilles et al. (2006) ARP = .23 Pos. = .04 Ham (2003), Ham 316 High alcohol 38% 90% In person Yes IAS Month RAPI 6 months n/a n/a AU FQ = −.19 and Hope (2005) use/problems ARP = −.06 Ham (2003), Ham 228 Low alcohol 39% 91% In person Yes Comp. of SPAI/IAS Month RAPI 6 months n/a CEOA AU FQ = −.17 and Hope (2006) use/problems ARP = −.09 Pos. = −.06 Neg. = .09 Ham (2009) 610 None 69% 13% In person Yes SIAS/SPS total Not rep. AUDIT Not rep. n/a AESES AU FQ = −.19 and DEQ ARP = .04 Pos. = .14 2701 Neg. = .03 (continued on next page) 2702

Appendix(continued) A (continued) Report information Sample size Rec./sel. % Female % Caucasian Data coll. Pub./ SA meas. AU ARP meas. ARP Motive AOEs meas. Correlations method peer-review time-frame time-frame meas. report

Ham, Bonin, 239 Non-drinkers excluded 59% 90% In person Yes SPAI Month RAPI 6 months DMQ n/a AU Quant. = −.23 and Hope (2007) AU Freq. = −.13 ARP = −.07 Cope = .10 Conf. = .05 Enh. = −.06 Soc. = .01 Ham, Casner, Bacon, 62 High alcohol 31% 81% In person Yes SIAS/SPS total Week RAPI 6 months n/a n/a AU Quant. = −.02 and Shaver (2011) use/problems AU Freq. = .02 ARP = .34 ..Shy ..Wie/AdcieBhvos3 21)2690 (2013) 38 Behaviors Addictive / White S.W. Schry, A.R. Ham, Zamboanga, 377 None 70% 77% Other Yes SIAS Not rep. AUDIT Not rep. n/a CEOA AU FQ = .03 and Bacon (2011) ARP = .11 Pos. = .23 Neg. = .22 Ham et al. (2009) 817 Non-drinkers excluded 74% 74% Other Yes SIAS Not rep. AUDIT Not rep. DMQ n/a AU FQ = −.08 ARP = .02 Cope = .28 Conf. = .30 Enh. = .12 Soc. = .13 Ham, Zamboanga, 715 Non-drinkers excluded 75% 68% In person Yes SIAS n/a n/a n/a n/a CEOA Pos. = .15 Olthuis, Casner, and Bui (2010) Heideman (2008) 14 High SA; high alcohol 44% 89% In person No SIAS and BFNE Month RAPI Month n/a n/a AU Quant. = −.24 use/problems AU Freq = .39 AU FQ = .11 ARP = −.20 LaBrie, Hummer, and 1168 None 70% 66% Other Yes SA subscale of SCS Month n/a n/a n/a n/a AU FQ = −.10 Neighbors (2008) Leonard and 86 Non-drinkers excluded 0% 92% Not rep. Yes FNE, SAD, and SA n/a n/a n/a n/a AEQ Pos. = .28 – Blane (1988) subscale of SCS 2706 Lewis et al. (2008) 316 High alcohol 54% 99% Other Yes IAS and LSAS Month RAPI and YAAPST Year DMQ n/a AU Quant. = −.09 use/problems AU FQ = −.10 ARP = .14 Cope = .28 Conf. = .24 McKeon (2005) 697 None 50% 90% Other No FNE and SAD Month RAPI Year n/a CEOA and AU Quant. = −.08 A-AESES AU Freq. = −.09 AU Binge = −.12 ARP = −.12 Pos. = .07 Neg. = .13 Merrill (2000) 264 Non-drinkers excluded 100% Not rep. In person No FNE 6 months sMAST Lifetime n/a n/a AU Quant. = .00 AU Freq. = −.06 ARP = .05 Mignogna (2010) 140 High alcohol 44% 85% In person No SIAS and FNE Month B-YAACQ Not rep. n/a n/a AU Quant. = −.01 use/problems and CAPS-r AU FQ = .01 ARP = .31 Miller (1989) 365 None 66% 70% In person No SAD Not rep. Drinking Conseq. Not rep. n/a n/a AU FQ = −.21 Scale of Drinking ARP = −.18 Styles Quest. Report information

Sample size Rec./sel. % Female % Caucasian Data coll. Pub./ SA meas. AU ARP meas. ARP Motive AOEs meas. Correlations method peer-review time-frame time-frame meas. report

Norberg et al. (2009), 118 High SA 61% 40% In person Yes SIAS/SPS 3 months RAPI and 3 months DMQ n/a AU Quant. = −.01 (2010), (2011) total and FNE modiﬁed TLFB AU Freq. = −.07 AU FQ = −.12 ARP = .10 Cope = .40 Conf. = .25 Enh. = .04 Soc. = .12 O'Grady, Cullum, 476 Non-drinkers excluded 52% 86% In person Yes SA subscale Month n/a n/a n/a n/a AU Quant. = −.04 Armeli, and of SCS Tennen (2011) O'Hare (1988), (1990) 606 None 56% 67% Other Yes Social Month Rutgers Survey Year n/a AEQ AU Quant. = −.14 Anxiety Scale ARP = −.04 Pos. = .15 Owen (1995) 312 None 50% 90% In person No FNE Week Problems Not rep. n/a n/a AU FQ = −.09 and SAD from Drinking ARP = −.05 Scale of SAQ Popali-Lehane (2006) 215 None 78% 43% In person No SIAS/SPS Year Core Alcohol Year DMQ CEOA AU Freq. = −.07 total and Drug Survey ARP = −.04 ..Shy ..Wie/AdcieBhvos3 21)2690 (2013) 38 Behaviors Addictive / White S.W. Schry, A.R. Cope = .25 Conf. = .34 Enh. = .01 Soc. = .13 Pos. = .09 Neg. = .19 Rohsenow (1982) 36 High alcohol 0% Not rep. In person Yes Social and Test 3 months n/a n/a n/a n/a AU FQ = −.41 use/problems Anxiety Scale Schry (2011) 588 Non-drinkers excluded 100% 88% Other No SIAS/SPS Year Norberg et al. Year DMQ n/a AU Quant. = −.05 total and (2009) list of ARPs AU Freq. = −.13 SPAI-23 AU Binge = −.06 ARP = .07 Cope = .25 Conf. = .27 Enh. = .04 Soc. = .08 Schry and 412 Non-drinkers excluded 100% 88% Other No SPAI-23 Month n/a n/a n/a DAMS AU Quant. = −.04 White (2009) AU Freq. = −.15 AU Binge = −.08

Cope = .19 – Schry and 383 Non-drinkers excluded 100% 88% Other No SIAS/SPS total Month n/a n/a n/a n/a AU Quant. = −.02 2706 White (2010) and SPAI-23 AU Freq. = −.17 AU Binge = −.05 Schry and 674 Non-drinkers excluded 69% 85% Other No SIAS/SPS total Month Norberg et al. Month n/a n/a AU Quant. = −.08 White (2012) (2009) list of ARPs AU Freq. = −.15 AU FQ = −.14 ARP = .03 Schwartz 9024 None 73% 62% Other Yes SIAS Month n/a n/a n/a n/a AU Quant. = −.10 et al. (2011) AU Freq. = −.11 Storch, Bagner, 287 None 72% 72% In person Yes SAS-A n/a PAI Not rep. n/a n/a ARP = .12 Geffken, and Baumeister (2004) Storch, Storch, 388 None 68% 70% In person Yes SAS-A n/a PAI Not rep. n/a n/a ARP = .11 Killiany, and Roberti (2005) (continued on next page) 2703 2704

Strahan, Panayiotou, 824 None 61% Not rep. Not rep. Yes SIAS Not rep. AUDIT Not rep. n/a Alcohol AU Quant. = −.07 Clements, and Outcome AU Freq. = −.12 Scott (2011) Exp, Quest. AU Binge = −.06 Pos. = .09 Neg. = .17 Terlecki, Buckner, 70 High/low SA; 31% 86% In person Yes SPS Month RAPI Month n/a n/a AU Quant. = .34 Larimer, and high alcohol AU Freq. = .29 Copeland (2011), use/problems AU FQ = .34 (2012) ARP = .32 Tran, Anthenelli, 152 High/low SA; 49% 94% In person Yes SIAS Year n/a n/a n/a AEQ AU Quant. = .12 Smith, Corcoran, high alcohol and AESES Pos. = .25 and Rofey (2004) use/problems Tran, Haaga, 229 None 66% 82% In person Yes SIAS Month n/a n/a n/a AEQ AU Quant. = −.05 and Chambless and AESES AU Freq. = −.14 (1997) Pos. = .17

Note. Measures only included if results were available for that measure. AEQ = Alcohol Expectancy Questionnaire; AESES = Alcohol Expectancies for Social Evaluative Situations Scale; AU = Alcohol Use; AUDIT = Alcohol Use Disorders Identiﬁcation Test; BFNE = Brief Fear of Negative Evaluation Scale; B-YAACQ = Brief Young Adult Alcohol Consequences Questionnaire; CAPS-r = College Alcohol Problems Scale — Revised; CEOA = Comprehensive Effects of Alcohol Scale; Conf. = Conformity Motives;

Cope = Coping Motives; DAMS = Drinking for Anxiety Management Scale; Data Coll. Method = Data Collection Method; DCSAS = Drinking to Cope with Social Anxiety Scale; DEQ = Drinking Expectancy Questionnaire; DMQ = Drinking Motives Ques- – tionnaire or DMQ-Revised; EDA = Effects of Drinking Alcohol Scale; Enh. = Enhancement Motives; Exp. = Expectancies; FNE = Fear of Negative Evaluation Scale; FQ = Frequency by Quantity Index; Freq. = Frequency; IAS = Interaction Anxiousness 2706 Scale; LSAS = Liebowitz Social Anxiety Scale; Meas. = Measure; n/a = Not Applicable; Neg. = Negative AOEs; Not rep. = Not Reported; PAI = Personality Assessment Inventory; Pos. = Positive AOEs; Pub/Peer-Review = Published and Peer-Reviewed; Quant. = Quantity; Quest. = Questionnaire; RAPI = Rutgers Alcohol Problems Index; Rec./Sel. = Recruitment and Selection Criteria; SA = Social Anxiety; SAD = Social Avoidance and Distress Scale; SAQ = Student Alcohol Question- naire; SAS-A = Social Anxiety Scale for Adolescents; SCS = Self-Consciousness Scale; SIAS = Social Interaction Anxiety Scale; SIDS = Social Impressions while Drinking Scale; sMAST = Short Michigan Alcohol Screening Test; Soc. = Social Motives; SPAI = Social Phobia and Anxiety Inventory; SPS = Social Phobia Scale; TLFB = Timeline Followback; WMH-CIDI = World Mental Health Initiative Version of the Composite International Diagnostic Interview; YAAPST = Young Adult Alcohol Problems Screening Test. A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706 2705

1 References *Gilles, D. M., Turk, C. L., & Fresco, D. M. (2006). Social anxiety, alcohol expectancies, and self-efficacy as predictors of heavy drinking in college students. Addictive American Psychiatric Association (2000). Diagnostic and statistical manual of mental Behaviors, 31(3), 388–398. http://dx.doi.org/10.1016/j.addbeh.2005.05.020. disorders (Text Revision). (4th ed.)Washington, DC: Author. Gonzalez, V. M., Bradizza, C. M., & Collins, R. L. (2009). Drinking to cope as a statistical Babor, T. F., Higgins-Biddle, J. C., Saunders, J. B., & Monteiro, M. G. (2001). The alcohol mediator in the relationship between suicidal ideation and alcohol outcomes use disorders identification test: Guidelines for use in primary care (2nd ed.)Geneva, among underage college drinkers. Psychology of Addictive Behaviors, 23(3), 443–451. Switzerland: World Health Organization. http://dx.doi.org/10.1037/a0015543. Bacon, A. K., & Ham, L. S. (2010). Attention to social threat as a vulnerability Grant, B. F., Hasin, D. S., Blanco, C., Stinson, F. S., Chou, S. P., Goldstein, R. B., et al. (2005). to the development of comorbid social anxiety and alcohol use disorders: The epidemiology of social anxiety disorder in the United States: Results from the An avoidance-coping cognitive model. Addictive Behaviors, 35,925–939. national epidemiologic survey on alcohol and related conditions. The Journal of http://dx.doi.org/10.1016/j.addbeh.2010.06.002. Clinical Psychiatry, 66(11), 1351–1361. *Bagner,D.M.,Storch,E.A.,&Preston,A.S.(2007).Romanticrelational Grant, V. V., Stewart, S. H., & Mohr, C. D. (2009). Coping-anxiety and coping-depression aggression: What about gender? Journal of Family Violence, 22(1), 19–24. motives predict different daily mood-drinking relationships. Psychology of Addictive http://dx.doi.org/10.1007/s10896-006-9055-x. Behaviors, 23(2), 226–237. http://dx.doi.org/10.1037/a0015006. *Bella, T. T., & Omigbodun, O. O. (2009). Social phobia in Nigerian university students: *Ham, L. S. (2003). Incorporating social anxiety into a model of college student Prevalence, correlatesandco-morbidity.Social Psychiatry and Psychiatric Epidemiology, problematic drinking. Dissertation Abstracts International: Section B. The Sciences 44(6), 458–463. http://dx.doi.org/10.1007/s00127-008-0457-3. and Engineering, 3524 (2004). Benton, S. L., Schmidt, J. L., Newton, F. B., Shin, K., Benton, S. A., & Newton, D. W. (2004). *Ham, L. S. (2009). Positive social alcohol outcome expectancies, social anxiety, and College student protective strategies and drinking consequences. Journal of Studies hazardous drinking in college students. Cognitive Therapy and Research, 33(6), on Alcohol, 65(1), 115–121. 615–623. http://dx.doi.org/10.1007/s10608-009-9248-8. Buckner, J. D. (2011). Anxiety and substance use disorders: Identifying underlying *Ham,L.S.,Bonin,M.,&Hope,D.A.(2007).Theroleofdrinkingmotivesin mechanisms and encouraging behavioral change. Anxiety Disorders: A Tri-Annual social anxiety and alcohol use. Journal of Anxiety Disorders, 21(8), 991–1003. Report, 8(1), 1–10. http://dx.doi.org/10.1016/j.janxdis.2006.10.014. *Buckner, J. D., Ecker, A. H., & Proctor, S. L. (2011). Social anxiety and alcohol problems: *Ham,L.S.,Casner,H.G.,Bacon,A.K.,&Shaver,J.A.(2011a).Speeches,strangers, The roles of perceived descriptive and injunctive peer norms. Journal of Anxiety and alcohol use: The role of context in social stress response dampening. Disorders, 25, 631–638. http://dx.doi.org/10.1016/j.janxdis.2011.02.003. Journal of Behavior Therapy and Experimental Psychiatry, 42,462–472. *Buckner, J. D., & Heimberg, R. G. (2010). Drinking behaviors in social situations http://dx.doi.org/10.1016/j.jbtep.2011.04.004. account for alcohol-related problems among socially anxious individuals. Psychology Ham,L.S.,&Hope,D.A.(2003).Collegestudents and problematic drinking: A review of Addictive Behaviors, 24,640–648. http://dx.doi.org/10.1037/a0020968. of the literature. Clinical Psychology Review, 23(5), 719–759. http://dx.doi.org/ Buckner, J. D., Heimberg, R. G., & Schmidt, N. B. (2011). Social anxiety and 10.1016/s0272-7358(03)00071-0. marijuana-related problems: The role of social avoidance. Addictive Behaviors, *Ham, L. S., & Hope, D. A. (2005). Incorporating social anxiety into a model of 36(1–2), 129–132. http://dx.doi.org/10.1016/j.addbeh.2010.08.015. college student problematic drinking. Addictive Behaviors, 30(1), 127–150. *Buckner, J. D., & Matthews, R. A. (2012). Social impressions while drinking account http://dx.doi.org/10.1016/j.addbeh.2004.04.018. for the relationship between alcohol-related problems and social anxiety. Addictive *Ham, L. S., & Hope, D. A. (2006). Incorporating social anxiety into a model of college Behaviors, 37, 533–536. http://dx.doi.org/10.1016/j.addbeh.2011.11.026. problem drinking: Replication and extension. Psychology of Addictive Behaviors, *Buckner, J. D., Schmidt, N. B., & Eggleston, A. M. (2006). Social anxiety and problematic 20(3), 348–355. http://dx.doi.org/10.1037/0893-164x.20.3.348. alcohol consumption: The mediating role of drinking motives and situations. Ham, L. S., Stewart, S. H., Norton, P. J., & Hope, D. A. (2005). Psychometric assessment of Behavior Therapy, 37(4), 381–391. http://dx.doi.org/10.1016/j.beth.2006.02.007. the Comprehensive Effects of Alcohol Questionnaire: Comparing a brief version to Buckner, J. D., Schmidt, N. B., Lang, A. R., Small, J. W., Schlauch, R. C., & Lewinsohn, P. M. the original full scale. Journal of Psychopathology and Behavioral Assessment, 27(3), (2008a). Specificity of social anxiety disorder as a risk factor for alcohol 141–158. http://dx.doi.org/10.1007/s10862-005-0631-9. and cannabis dependence. Journal of Psychiatric Research, 42(3), 230–239. *Ham, L. S., Zamboanga, B. L., & Bacon, A. K. (2011b). Putting thoughts into context: http://dx.doi.org/10.1016/j.jpsychires.2007.01.002. Alcohol expectancies, social anxiety, and hazardous drinking. Journal of Cognitive Buckner,J.D.,Timpano,K.R.,Zvolensky,M.J.,Sachs-Ericsson,N.,&Schmidt,N.B. Psychotherapy, 25(1), 47–60. http://dx.doi.org/10.1891/0889-8391.25.1.47. (2008b). Implications of comorbid alcohol dependence among individuals *Ham, L. S., Zamboanga, B. L., Bacon, A. K., & Garcia, T. A. (2009). Drinking motives as me- with social anxiety disorder. Depression and Anxiety, 25(12), 1028–1037. diators of social anxiety and hazardous drinking among college students. Cognitive http://dx.doi.org/10.1002/da.20442. Behaviour Therapy, 38(3), 133–145. http://dx.doi.org/10.1080/16506070802610889. Buckner, J. D., & Turner, R. J. (2009). Social anxiety disorder as a risk factor for alcohol use *Ham, L. S., Zamboanga, B. L., Olthuis, J. V., Casner, H. G., & Bui, N. (2010). No disorders: A prospective examination of parental and peer influences. Drug and Alcohol fear, just relax and play: Social anxiety, alcohol expectancies, and drinking Dependence, 100(1), 128–137. http://dx.doi.org/10.1016/j.drugalcdep.2008.09.018. games among college students. Journal of American College Health, 58(5), *Burke, R. S., & Stephens, R. S. (1997). Effect of anxious affect on drinking 473–479. http://dx.doi.org/10.1080/07448480903540531. self-efficacy in college students. Psychology of Addictive Behaviors, 11(1), Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. San Diego, CA: 65–75. http://dx.doi.org/10.1037/0893-164x.11.1.65. Academic Press. Chambless, D. L., Hunter, K., & Jackson, A. (1982). Social anxiety and assertiveness: A *Heideman, P. W. (2008). Combining cognitive behavioral therapy with an alcohol comparison of the correlations in phobic and college student samples. Behaviour intervention to reduce alcohol problems among socially anxious college students. Research and Therapy, 20(4), 403–404. http://dx.doi.org/10.1016/0005-7967(82) Dissertation Abstracts International: Section B. The Sciences and Engineering, 6413 90101-2. (2009). *Clerkin, E. M., & Barnett, N. (2012). The separate and interactive effects of drinking mo- Higgins, R. L., & Marlatt, G. A. (1975). Fear of interpersonal evaluation as a determinant tives and social anxiety symptoms in predicting drinking outcomes. Addictive Behav- of alcohol consumption in male social drinkers. Journal of Abnormal Psychology, iors. http://dx.doi.org/10.1016/j.addbeh.2012.01.005 (Advance online publication). 84(6), 644–651. http://dx.doi.org/10.1037/0021-843x.84.6.644. Cooper, M. L. (1994). Motivations for alcohol use among adolescents: Development Hingson, R., Heeren, T., Winter, M., & Wechsler, H. (2005). Magnitude of alcohol- and validation of a four-factor model. Psychological Assessment, 6(2), 117–128. related mortality and morbidity among U.S. college students ages 18–24: http://dx.doi.org/10.1037/1040-3590.6.2.117. Changes from 1998 to 2001. Annual Review of Public Health, 26,259–279. Cooper, H. (2010). Research synthesis and meta-analysis: A step-by-step approach. http://dx.doi.org/10.1146/annurev.publhealth.26.021304.144652. Thousand Oaks, CA: SAGE Publications. Holroyd, K. A. (1978). Effects of social anxiety and social evaluation on beer consumption Cooper, M. L., Frone, M. R., Russell, M., & Mudar, P. (1995). Drinking to regulate positive and and social interaction. Journal of Studies on Alcohol, 39(5), 737–744. negative emotions: A motivational model of alcohol use. Journal of Personality and Social Johnston, L. D., O'Malley, P. M., Bachman, J. G., & Schulenberg, J. E. (2010). Monitoring Psychology, 69(5), 990–1005. http://dx.doi.org/10.1037/0022-3514.69.5.990. the future: National survey results on drug use, 1975–2009: Volume I, Secondary *Dhuse, S. R. (2005). Consequences of binge drinking: Risk and protective factors. Dis- school students. (NIH Publication No. 10-7584). Bethesda, MD: National Institute sertation Abstracts International: Section B. The Sciences and Engineering, 2006, 537. on Drug Abuse. *Eggleston, A. M., Woolaway-Bickel, K., & Schmidt, N. B. (2004). Social anxiety and alcohol Jones, B. T., Corbin, W., & Fromme, K. (2001). A review of expectancy theory and use: Evaluation of the moderating and mediating effects of alcohol expectancies. Journal alcohol consumption. Addiction, 96(1), 57–72. of Anxiety Disorders, 18(1), 33–49. http://dx.doi.org/10.1016/j.janxdis.2003.07.005. Kidorf, M., & Lang, A. R. (1999). Effects of social anxiety and alcohol expectancies Falk,D.E.,Yi,H.-y,&Hilton,M.E.(2008).Ageofonsetandtemporalsequencing on stress-induced drinking. Psychology of Addictive Behaviors, 13(2), 134–142. of lifetime — DSM-IV — Alcohol use disorders relative to comorbid http://dx.doi.org/10.1037/0893-164x.13.2.134. mood and anxiety disorders. Drug and Alcohol Dependence, 94(1–3), 234–245. Kushner, M. G., & Sher, K. J. (1993). Comorbidity of alcohol and anxiety disorders http://dx.doi.org/10.1016/j.drugalcdep.2007.11.022. among college students: Effects of gender and family history of alcoholism. Addictive Fromme, K., Stroot, E. A., & Kaplan, D. (1993). Comprehensive effects of alcohol: Behaviors, 18(5), 543–552. http://dx.doi.org/10.1016/0306-4603(93)90070-p. Development and psychometric assessment of a new expectancy questionnaire. *LaBrie, J. W., Hummer, J. F., & Neighbors, C. (2008). Self-consciousness moderates Psychological Assessment, 5(1), 19–26. http://dx.doi.org/10.1037/1040-3590.5.1.19. the relationship between perceived norms and drinking in college stu- *Gilles, D. M. (2005). Social anxiety, alcohol expectancies, and self-efficacy as dents. Addictive Behaviors, 33(12), 1529–1539. http://dx.doi.org/10.1016/ predictors of heavy drinking in college students. Dissertation Abstracts Internation- j.addbeh.2008.07.008. al: Section B. The Sciences and Engineering, 5401 (2007). LaBrie,J.,Pedersen,E.R.,Neighbors,C.,&Hummer,J.F.(2008).Theroleofself-consciousness in the experience of alcohol-related consequences among college students. Addictive Behaviors, 33(6), 812–820. http://dx.doi.org/10.1016/j.addbeh.2008.01.002. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical – 1 * Indicates reports included in the meta-analysis. data. Biometrics, 33(1), 159 174. 2706 A.R. Schry, S.W. White / Addictive Behaviors 38 (2013) 2690–2706

*Leonard, K. E., & Blane, H. T. (1988). Alcohol expectancies and personality Sayette, M. A. (1993). An appraisal-disruption model of alcohol's effects on stress re- characteristics in young men. Addictive Behaviors, 13(4), 353–357. http://dx.doi.org/ sponses in social drinkers. Psychological Bulletin, 114(3), 459–476. http://dx.doi.org/ 10.1016/0306-4603(88)90041-x. 10.1037/0033-2909.114.3.459. LeSure-Lester, G. E. (2001). Dating competence, social assertion and social anxiety *Schry, A. R. (2011). Social anxiety as a risk factor of sexual victimization in college among college students. College Student Journal, 35(2), 317–320. women (Unpublished master's thesis). Virginia Tech, Blacksburg, VA. *Lewis, M. A., Hove, M. C., Whiteside, U., Lee, C. M., Kirkeby, B. S., Oster-Aaland, L., et al. Schry, A. R., Roberson-Nay, R., & White, S. W. (2012). Measuring social anxiety in (2008). Fitting in and feeling fine: Conformity and coping motives as mediators of college students: A comprehensive evaluation of the psychometric properties of the the relationship between social anxiety and problematic drinking. Psychology of SPAI-23. Psychological Assessment, 24,846–854. http://dx.doi.org/10.1037/a0027398. Addictive Behaviors, 22(1), 58–67. http://dx.doi.org/10.1037/0893-164x.22.1.58. *Schry, A. R., & White, S. W. (2009). [Social anxiety and sexual victimization: Fall 2009]. Lipsey, M. W., & Wilson, D. B. (2001). Practical Meta-Analysis, vol. 49, Thousand Oaks, Unpublished raw data. CA: Sage Publications, Inc. *Schry, A. R., & White, S. W. (2010). [Social anxiety and sexual victimization: Spring Mallett, K. A., Marzell, M., Varvil-Weld, L., Turrisi, R., Guttman, K., & Abar, C. (2011). 2010]. Unpublished raw data. One-time or repeat offenders? An examination of the pattern of alcohol-related *Schry, A. R., & White, S. W. (2012). [Social anxiety and alcohol use: Spring 2012]. consequences experienced by college students across the freshman year. Addictive Unpublished raw data. Behaviors, 36, 508–511. http://dx.doi.org/10.1016/j.addbeh.2010.12.022. *Schwartz,S.J.,Beyers,W.,Luyckx,K.,Soenens,B.,Zamboanga,B.L.,Forthun,L.F.,etal. Mattick, R. P., & Clarke, J. C. (1998). Development and validation of measures of social (2011). Examining the light and dark sides of emerging adults' identity: A study of iden- phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, tity status differences in positive and negative psychosocial functioning. Journal of Youth 36(4), 455–470. http://dx.doi.org/10.1016/s0005-7967(97)10031-6. and Adolescence, 40(7), 839–859. http://dx.doi.org/10.1007/s10964-010-9606-6. *McKeon, N. D. (2005). The role of alcohol expectancies in the relationship between alco- Steele, C. M., Southwick, L., & Pagano, R. (1986). Drinking your troubles away: The role hol use and social anxiety in college students: A social–cognitive framework. Disser- of activity in mediating alcohol's reduction of psychological stress. Journal of Ab- tation Abstracts International: Section B. The Sciences and Engineering, 3418 (2005). normal Psychology, 95(2), 173–180. http://dx.doi.org/10.1037/0021-843x.95.2.173. *Merrill, K. A. (2000). Effects of anxiety and anxiety sensitivity on alcohol use in fe- Stewart, S. H., Morris, E., Mellings, T., & Komar, J. (2006). Relations of social anxiety males. Dissertation Abstracts International: Section B. The Sciences and Engineering, variables to drinking motives, drinking quantity and frequency, and alcohol- 3852 (January 2001). related problems in undergraduates. Journal of Mental Health, 15(6), 671–682. *Mignogna, J. (2010). A test of moderating factors of brief interventions for hazardous http://dx.doi.org/10.1080/09638230600998904. alcohol use among college students. Dissertation Abstracts International: Section B. *Storch, E. A., Bagner, D. M., Geffken, G. R., & Baumeister, A. L. (2004). Association between The Sciences and Engineering, 6446 (2011). overt and relational aggression and psychosocial adjustment in undergraduate *Miller, P. M. (1989). A theory for alcohol use: A general-specific factor risk model. Disser- college students. Violence and Victims, 19(6), 689–700. http://dx.doi.org/ tation Abstracts International: Section B. The Sciences and Engineering, 4061 (February 10.1891/vivi.19.6.689.66342. 1991). *Storch, E. A., Storch, J. B., Killiany, E. M., & Roberti, J. W. (2005). Self-reported psycho- Morris, E. P., Stewart, S. H., & Ham, L. S. (2005). The relationship between social anxiety pathology in athletes: A comparison of intercollegiate student-athletes and disorder and alcohol use disorders: A critical review. Clinical Psychology Review, non-athletes. Journal of Sport Behavior, 28(1), 86–98. 25(6), 734–760. http://dx.doi.org/10.1016/j.cpr.2005.05.004. *Strahan, E. Y., Panayiotou, G., Clements, R., & Scott, J. (2011). Beer, wine, and social anx- National Institute on Alcohol Abuse and Alcoholism, Task Force on Recommended Alcohol iety: Testing the “self-medication hypothesis” in the US and Cyprus. Addiction Re- Questions (October 15–16). Task Force on Recommended Alcohol Questions–National search & Theory, 19(4), 302–311. http://dx.doi.org/10.3109/16066359.2010.545152. Council on Alcohol Abuse and Alcoholism recommended sets of alcohol consumption Substance Abuse and Mental Health Services Administration (2012). Results from the questions. Retrieved from. http://www.niaaa.nih.gov/Resources/ResearchResources/ 2011 National Survey on Drug Use and Health: Summary of national findings. TaskForce.htm (NSDUH Series H-44, HHS Publication No. (SMA) 12-4713). Rockville, MD: Sub- Nitka, D., & O'Connor, R. M. (2011). Social anxiety and reasons for predrinking. Poster stance Abuse and Mental Health Services Administration. presented at the meeting of Research Society on Alcoholism, Atlanta, GA. *Terlecki, M. A., Buckner, J. D., Larimer, M. E., & Copeland, A. L. (2011). The role of social *Norberg, M. M., Norton, A. R., & Olivier, J. (2009). Refining measurement in the study anxiety in a brief alcohol intervention for heavy-drinking college students. Journal of social anxiety and student drinking: Who you are and why you drink of Cognitive Psychotherapy, 25(1), 7–21. http://dx.doi.org/10.1891/0889-8391.25.1.7. determines your outcomes. Psychology of Addictive Behaviors, 23(4), 586–597. *Terlecki, M. A., Buckner, J. D., Larimer, M. E., & Copeland, A. L. (2012). Brief motivational in- http://dx.doi.org/10.1037/a0016994. tervention for college drinking: The synergistic impact of social anxiety and perceived *Norberg, M. M., Norton, A. R., Olivier, J., & Zvolensky, M. J. (2010). Social anxiety, drinking norms. Psychology of Addictive Behaviors. http://dx.doi.org/10.1037/a0027982 reasons for drinking, and college students. Behavior Therapy, 41(4), 555–566. (Advance online publication). http://dx.doi.org/10.1016/j.beth.2010.03.002. Thomas, S. E., Randall, C. L., & Carrigan, M. H. (2003). Drinking to cope in socially anx- *Norberg, M. M., Olivier, J., Alperstein, D. M., Zvolensky, M. J., & Norton, A. R. (2011). Adverse ious individuals: A controlled study. Alcoholism, Clinical and Experimental Research, consequences of student drinking: The role of sex, social anxiety, drinking motives. 27(12), 1937–1943. http://dx.doi.org/10.1097/01.alc.0000100942.30743.8c. Addictive Behaviors, 36, 821–828. http://dx.doi.org/10.1016/j.addbeh.2011.03.010. *Tran, G. Q., Anthenelli, R. M., Smith, J. P., Corcoran, K. J., & Rofey, D. L. (2004). Alcohol *O'Grady, M. A., Cullum, J., Armeli, S., & Tennen, H. (2011). Putting the relationship use, cognitive correlates of drinking and change readiness in hazardous drinkers between social anxiety and alcohol use into context: A daily diary investigation with high versus low social anxiety. Journal of Studies on Alcohol, 65(6), 715–724. of drinking in response to embarrassing events. Journal of Social and Clinical *Tran,G.Q.,Haaga,D.A.F.,&Chambless,D.L.(1997).Expectingthatalcohol Psychology, 30(6), 599–615. http://dx.doi.org/10.1521/jscp.2011.30.6.599. use will reduce social anxiety moderates the relation between social anxiety *O'Hare, T. M. (1988). Alcohol expectancies and social anxiety as predictors of alcohol and alcohol consumption. Cognitive Therapy and Research, 21(5), 535–553. consumption and problem drinking in college undergraduates. Dissertation http://dx.doi.org/10.1023/a:1021857402164. Abstracts International: Section B. The Sciences and Engineering, 3160 (April 1989). Wechsler, H., Lee, J. E., Kuo, M., Seibring, M., Nelson, T. F., & Lee, H. (2002). Trends *O'Hare, T. M. (1990). Alcohol expectancies and social anxiety in male and female in college binge drinking during a period of increased prevention efforts: Findings undergraduates. Addictive Behaviors, 15(6), 561–566. http://dx.doi.org/10.1016/ from 4 Harvard School of Public Health College alcohol study surveys: 1993–2001. 0306-4603(90)90057-5. Journal of American College Health, 50(5), 203–217. *Owen, D. L. (1995). A study of college student drinking patterns and the mediating White, H. R., & Labouvie, E. W. (1989). Towards the assessment of adolescent problem effects of gender identity and social anxiety. Dissertation Abstracts International: drinking. Journal of Studies on Alcohol, 50(1), 30–37. Section B. The Sciences and Engineering, 3513 (March 1996). Wilson, D. B. (2010). Meta-analysis macros for SAS, SPSS, and Stata. Retrieved from. Park, C. L., & Levenson, M. R. (2002). Drinking to cope among college students: Preva- http://mason.gmu.edu/~dwilsonb/ma.html lence, problems and coping processes. Journal of Studies on Alcohol, 63(4), 486–497. Zamboanga, B. L., & Ham, L. S. (2008). Alcohol expectancies and context-specific *Popali-Lehane, L. (2006). Predicting alcohol problem use in college students with social drinking behaviors among female college athletes. Behavior Therapy, 39(2), phobia. Dissertation Abstracts International: Section B. The Sciences and Engineering, 162–170. http://dx.doi.org/10.1016/j.beth.2007.06.002. 4696 (2007). Zamboanga, B. L., Schwartz, S. J., Ham, L. S., Borsari, B., & Van Tyne, K. (2010). Alcohol *Rohsenow, D. J. (1982). Social anxiety, daily moods, and alcohol use over time among expectancies, pregaming, drinking games, and hazardous alcohol use in a heavy social drinking men. Addictive Behaviors, 7(3), 311–315. http://dx.doi.org/ multiethnic sample of college students. Cognitive Therapy and Research, 34(2), 10.1016/0306-4603(82)90062-4. 124–133. http://dx.doi.org/10.1007/s10608-009-9234-1.